Gradient
Lesson 5.9ProjectModelCode~2 h

Project: a Bayesian spam filter

Build a real spam filter from word counts and Bayes’ rule, and measure how well it works on emails it never saw.

Module 5 · 0/9 done

Project: a spam filter

Early email spam filters used exactly the maths from this module, and they worked very well. You’ll build one from scratch and test it on emails it has never seen.

  1. Train: count words in spam and in normal (“ham”) email.
  2. Classify: combine the evidence from every word with Bayes’ rule, in log space.
  3. Evaluate: measure accuracy on held-out emails, and see which words give spam away.